Subject: Re: seemingly dismal performance of NetBSD-1.1A/sun3 file I/O....
To: Justin T. Gibbs <gibbs@freefall.freebsd.org>
From: John F. Woods <jfw@funhouse.com>
List: current-users
Date: 04/12/1996 12:06:41
>>The makefiles contain information that allows you to just ftp the differences
>>from the freebsd site, then it will automatically contact a host which hasn't
>>had the corresponding version of the original source for a couple of years...
>Seeing as the FreeBSD ports collection hasn't even existed for "a couple of
>years", this is blatently false.

Unfortunately, I don't remember which package(s) it was that I had this
problem with, so I can't easily uncover the evidence; it is also possible
that I picked up the package from somewhere other than ftp.freebsd.org,
in which case FreeBSD's collections may well be all up to date (is there,
by any chance, a daemon that checks each Makefile in the ports directory?).
But I most certainly ran across the above mentioned problem.

> The makefile usually tries a list of sites, including ftp.freeBSD.org

The makefile for ytalk on ftp.freebsd.org refers solely to knot.queensu.ca;
perhaps there is a reference to ftp.freebsd.org in FreeBSD's master makefile
as a site of last resort, but it certainly isn't present in that makefile on
ftp.freebsd.org.  (That *is*, at least, a valid source of ytalk, so this
isn't the package I had trouble with.)

> If for some reason you still can't get the tarball for the port, your
> sources for the ports collection must be really old and you should update
> them.

I don't have any "sources" for the ports collection, unless you mean the list
of hosts I try when looking for sources (usually I consult archie, which has
its own problems with stale data).  The problem is that someone *else* has
to update an archive (somewhere), and it is much too easy for one of these
embedded links to become stale without notice to an archive maintainer.
Note that the same problem is widespread on the WorldWide Web.

When I had the freebsd ports scheme described to me, I thought it was really
cool.  Unfortunately, it let me down when I finally tried to make use of it,
in a way that exposed an important design deficiency.  Considering the benefit
of not *having* to spend disk space replicating all of these original source
packages, it's probably a deficiency that's quite reasonable to live with;
that doesn't keep it from being a hole in the design, however.